基于Lasso回归的红皮病性银屑病伴发代谢综合征风险因素分析及预测模型构建  

Risk factors analysis and prediction model construction of metabolic syndrome in erythrodermic psoriasis based on lasso regression

作　　者：赵心源 邹玥敏 周冬梅[2] ZHAO Xinyuan;ZOU Yuemin;ZHOU Dongmei(Graduate school Beijing University of Chinese Medicine,Beijing 100029,China)

机构地区：[1]北京中医药大学研究生院,北京100029 [2]首都医科大学附属北京中医医院皮肤科

出　　处：《实用皮肤病学杂志》2024年第1期12-16,24,共6页Journal of Practical Dermatology

摘　　要：目的探究红皮病性银屑病患者发生代谢综合征的危险因素,构建可用于临床的预测模型。方法选取2014年3月至2022年6月北京中医医院皮肤科住院患者的病历数据,包括人口学、实验室资料及结局指标等,以是否发生代谢综合征为因变量,采用Logistic回归及Lasso回归筛选出最优潜在预测因子,建立风险预测列线图。通过受试者工作特征曲线(ROC曲线)评估模型区分度,Hosmer Lemeshow检验和Calibration曲线评估模型的一致性和获益性。结果以高血压、甘油三酯、糖尿病、高密度脂蛋白胆固醇、低密度脂蛋白胆固醇、非酒精性脂肪肝作为红皮病性银屑病患者发生代谢综合征预测指标,并建立列线图,评估模型得到训练集AUC=0.960,验证集AUC=0.942,显示模型具有较高预测能力。Hosmer Lemeshow检验显示χ^(2)=0.82,P=0.9997(P>0.1),calibration曲线所得AUC>0.9,说明模型区分度、准确度及可信度均良好。结论基于Lasso回归建立的临床预测模型具有良好的区分度、拟合度以及临床有效性,可应用于临床,为防治红皮病性银屑病患者合并代谢综合征提供科学依据。Objective To investigate the risk factors associated with metabolic syndrome(MS)in patients with erythrodermic psoriasis and develop a predictive model for clinical application.Methods The medical records of inpatients admitted to the dermatology ward of Beijing Hospital of Traditional Chinese Medicine from March 2014 to June 2022 were selected,including demographic data,laboratory data and outcome indicators.Logistic regression and Lasso regression were used to screen out the optimal potential predictors based on the occurrence of metabolic syndrome,and a risk prediction nomogram was constructed.The model discrimination was assessed using the receiver operating characteristic(ROC)curve,while its consistency and utility were evaluated through the Hosmer Lemeshow test and calibration curve.Results Hypertension,triglycerides,diabetes,high-density lipoprotein,low-density lipoprotein,and non-alcoholic fatty liver were used as predictors for MS in patients with erythrodermic psoriasis to establish a nomogram.The AUC value was 0.960 in the training set and 0.942 in the validation set,indicating that the model had high predictive ability.Hosmer Lemeshow test showed thatχ^(2)=0.82,P=0.9997(P>0.1),and the AUC value of calibration curve was 0.942(AUC>0.9),indicating that the model has excellent discrimination,accuracy and credibility.Conclusion The clinical prediction model established in this study exhibits strong discrimination,fitting and clinical effectiveness,which can be applied in clinical practice and provide a scientific basis for the prevention and treatment of MS in patients with erythrodermic psoriasis.

关 键 词：银屑病 红皮病性 代谢综合征 Lasso回归 临床预测模型 

分 类 号：R758.63[医药卫生—皮肤病学与性病学]